Measuring the Structure of a Technology System for Directing Technological Transition

Abstract Technological advancements have generated a “techno‐sphere” within which all humans live. However, the capacity to direct technology development lags far behind technology development itself. This study deciphers the structural characteristics of a technology system using three pairs of features: systemicity and complexity (scalar), centrality and diversity (structural), and adaptability and inertia (structural); and at micro‐, meso‐, and macrolevels. By applying this approach in Chinese agricultural and water technology systems in the Yellow River Region and the Yangtze River Region from the beginning of agriculture in ≈8000 BC to the end of preindustrial agriculture in 1911, it is found that there exist trade‐off relationships between the centrality and diversity of a technology system, there exist alternative dominations of adaptivity and inertia in development of a technology system, and there exist time‐lag phenomena of change in a technology system between mesolevel and macrolevel. It is also identified that a larger‐scale, more diverse and adaptive technology system is observed in the Yellow River Region whereas the technology system in the Yangtze River Region is more rapidly expanding in scale and mainly dominated by inertia. These discoveries will assist increasing the capacity of managing and directing technological transition in future.

technologies that mobilize land and water resources. Fishery, forestry, animal grazing and husbandry, which were carried out on a limited scale in ancient China [5], were beyond the scope of this study.

Data collection
Content analysis, a well-tested method to systematically extract unstructured textural information for a wide range of studies [6,7], was adopted to extract technology information from the data sources.
The technology information variables were listed in Table S1. The first three variables provided the basic information on the names of specific technologies, and the time and locations where these technologies were invented, developed, or implemented. The final variable was designed to establish connections among technologies, which were based on the evolutionary information between technologies as described in encyclopedias. Two technologies were connected if one evolved based on the other, or where one has influenced/inspired the development of the other. Some examples of connections between technologies (underlined) include: Leisi is a primitive agricultural tool, upon which shovels, sickles and plows were developed; emergence of shovels marked the beginning of cultivation practices; and iron tools for agricultural development were facilitated by iron-making techniques. Establishment of the technology network The agricultural and water technologies extracted from the selected encyclopedias were firstly categorized based on the Chinese Classified Thesaurus (CCT), which is currently the most comprehensive linguistic system used for classification of words in Chinese [8]. As adapted in a previous study [9], a hierarchical structure was established for different technologies based on their functionalities (Table S2). The classification was mutually exclusive, which meant that one technology can be classified into only one group at any level.
A square matrix was then built to store the connection information between any two technologies to establish a technology network. Each technology was represented by a node, and the influential relationships were represented by edges in the technology network. The technology network was developed based on the following principles: (a) Cumulativeness: all technologies from previous periods were accumulated unless there were specific mentions that they were abandoned, as new technologies were very often adopted to complement existing ones [10]; (b) Interdependency: when connection exists, technologies within the same historical period were assumed to be inter-influencing with each other, i.e. two-directional connections, as inevitable knowledge exchanges among technology users [11]; (c) Uni-directionality: As time is irreversible, technology inheritance can only exist from earlier periods to the later one, but not vice versa.
Both the technology information and connections were cross-checked by two independent coders during the manual coding process to eliminate potential perceptual biases. The Krippendorff's alpha [12] was used to determine the degree of disagreement among the coders, which was kept above 80% as recommended by Poindexter and McCombs [13]. For any node d (as individual technology) in the network (equation (1) to (4) The above values were calculated for each d (each technology in the technology network). The centrality feature was calculated as normalised degree value, and the diversity feature was calculated as normalised clustering coefficient value.
For the adaptability and inertia features, normalised degree (D), closeness (C) and betweenness (B) to measure the level of influences of technologies that are newly developed in a period and those that are inherited from previous periods, respectively. Assume a technology system at a historical period (t) with total number of technologies N, and the number of existing technologies K, which are inherited from previous period (t-1) (equation (6) to (7)): For an individual technology : For an individual technology : This equation combines the effects of the three indicators at micro-scale to differentiate the comprehensive influences from the legacy technology sub-system and that from the innovative technology sub-system within a network. Due to the accumulated nature of a technology system, we measure inertia for the legacy technology sub-system by the changes of influences from previous periods. A power function is chosen to express the different contribution of the three indicators in their comprehensive effect. This is because the importance of degree, closeness and betweenness is ranked in a descending order, based on their functional characteristics in a network [14]. Similarly, the key technological sub-systems are formed by individual technologies with high inertia and/or high adaptability. The technology sub-systems with both low inertia and low adaptability are considered with limited influences on the entire network behaviours.
Ordinary Least Square (OLS) linear regression analysis and piece-wise linear regression were conducted using R (version 3.5.0, on platform x86_64, apple darwin15.6.0) base function "lm" and package "segmented" (https://cran.r-project.org/web/packages/segmented/index.html), which calculates the least squares fit in linear model using the QR decomposition method and estimates the break points between two linear regression segments using a score-based approach [15].
The sample sizes (n) for each historical period were: For the Adaptability-Inertia feature, technological sub-system in each historical period was divided among those that were inherited from previous periods (legacy technologies) and those that were newly developed in the current period (innovative technologies). The key sub-system included those technologies with greater than 0 normalised adaptability and/or inertia. A threshold of 0 was to ensure any technology with change of network measure are included. Figure S1 illustrates the distributions of different types of agricultural and water technologies in terms of their centrality (horizontal axis) and diversity (vertical axis) values for the Yellow River Region and the Yangtze River Region in each historical period. It was observed that the primitive plows ("Leisi"), shovels, and hoes from Agricultural Engineering formed the initial key sub-system from the Figure S1 Evolution of relationships between centrality and diversity for both river regions in time Figure S2 illustrates the distributions of different types of agricultural and water technologies in terms of their adaptability (plus symbol) and inertia (dot symbol) values for the Yellow River Region and the Yangtze River Region in each historical period. It was shown that Agricultural Engineering technologies (e.g. tangible tools and irrigation infrastructures) were identified to have high adaptability and inertia during the early Neolithic to WJ Period, especially for the Yellow River

B. The evolution of centrality and diversity values of technology in time for both regions
Region. In the succeeding periods, increasing numbers of technologies related to farming procedures (Agricultural Practice) and theoretical understandings (Agricultural Theory) were more effective to maintain past knowledge, as well as to stimulate development of innovative technologies for both river regions.